1. Field of the Invention
The present invention relates to a robotics apparatus including a control unit for controlling a first driving source for driving a first link, a second driving source for driving a second link, and a third driving source for driving both the first link and the second link, and to a robot control method, a program, and a recording medium.
2. Description of the Related Art
In a method for controlling a manipulator, it becomes more and more important for a hand tip to be able to flexibly come into contact with an object. If the flexible contact of the hand tip with the object is applied to industrial robots, collaborative work between a human and a robot is realized, or component fitting work is facilitated by controlling a direction of flexibility of the hand tip.
If the flexible contact described above is applied to leg-walking type mobile robots, legs can softly come into contact with a ground surface to reduce a shock applied on a trunk, or the robot can stably walk on an irregular terrain by absorbing a level difference.
In order to realize the control of flexibility of the hand tip, impedance control using a force sensor mounted on the hand tip, or control using an artificial muscle actuator is performed. It is known that a human muscle is an actuator as well as a control mechanism having a variable viscoelasticity. Among the artificial muscle actuators, in particular, a pneumatic rubber artificial muscle actuator as represented by a McKibben artificial muscle actuator has viscoelastic properties similar to those of the muscles. Therefore, by controlling softness of the artificial muscle actuator provided to the manipulator, the hand tip can come into contact with the object with arbitrary flexibility.
In the document, “Mechanical Properties of Robot Arm Operated with Muscle Coordinate System Consisted of Bi-articular Muscles and Mono-articular Muscles, by Toru Oshima, Tomohiro Fujikawa, and Minayori Kumamoto, Journal of the Japan Society of Precision Engineering, Vol. 66, No. 1, pp. 141 to 146”, it is proposed, that a 3-pair 6-muscle manipulator including a bi-articular simultaneous driving actuator for simultaneously driving a first link and a second link in addition to artificial muscle actuators for respectively driving the first link and the second link. Oshima et. al teaches features in the following stiffness properties. Specifically, when elasticities of the artificial muscle actuators are set equal to each other, a direction of an external force applied onto the hand tip and a direction of movement become the same in the case where the external force is applied to the hand tip on a line connecting a first joint and the hand tip. It is difficult to use the McKibben actuator for motion control because the McKibben actuator has nonlinearity in viscoelastic properties and control is required to be performed with antagonistic arrangement.
On the other hand, Japanese Patent Application Laid-Open No. 2012-86354 teaches that a simple feedback control system using saturation of a control input is derived to control a joint angle in addition of stiffness control for the hand tip in a simultaneous manner. However, the 3-pair 6-muscle manipulator requires six driving sources to be provided. Therefore, for example, when the McKibben artificial muscle actuator is used, a mechanical configuration including a control valve becomes complex.
On the other hand, in “Motor Development of an Pneumatic Musculoskeletal Infant Robot, by Kenichi Narioka and Koh Hosoda, 2011, IEEE International Conference on Robotics and Automation, pp. 963 to 968, 2011”, artificial muscle actuators are used for lower limbs of an infant robot. By acquiring a contractile-force command value by learning, a crawling motion is realized. In Narioka et. al, one actuator is provided to each joint in a direction of folding the first link and the second link with respect to the manipulator, whereas a single bi-articular simultaneous driving actuator is provided to unfold the two links.
In Narioka et. al, a learning control system is configured based on neural oscillators to acquire the contractile-force command value for realizing the crawling motion by linking motions of upper limbs and motions of the lower limbs.
In Narioka et. al, the realization of the crawling motion by a fullbody motion of the infant robot is considered, but positioning control or compliance control by the manipulator alone is not considered. Moreover, the learning control requires time to generate the control input. Therefore, it is difficult to apply the learning control to purposes of use in which a target trajectory frequently changes.